International Journal of Machine Tools & Manufacture 43 (2003) 1327–1336

Selection of optimum conditions for maximum material removal

rate with surface finish and damage as constraints in SiC grinding
Anne Venu Gopal, P. Venkateswara Rao ∗
Department of Mechanical Engineering, Indian Institute of Technology Delhi, New Delhi 110 016, India

Received 4 March 2003; received in revised form 3 June 2003; accepted 10 June 2003

Abstract

Efficient grinding of structural ceramics requires judicious selection of operating parameters to maximize removal rate while
controlling surface integrity. Grinding of silicon carbide is difficult because of its low fracture toughness, making it very sensitive
to cracking. In the present work, experiments were carried out to study the effect of wheel parameters; grain size and grain density
and grinding parameters; depth of cut and feed on the surface roughness and surface damage. The significance of the grinding
parameters on the selected responses was evaluated using analysis of variance. Mathematical models were developed using the
experimental data considering only the significant parameters. A genetic algorithm (GA) code has been developed to optimize the
grinding conditions for maximum material removal, using a multi-objective function model, by imposing surface roughness and
surface damage constraints. The choice of including manufacturer’s constraints on the basis of functional requirements of the
component for maximizing the production rate was also embedded in the GA code.
 2003 Elsevier Ltd. All rights reserved.

Keywords: Ceramic grinding; Surface roughness; Surface damage; Genetic algorithms

1. Introduction which are visible, and the later damage is due to median
and lateral cracks that are formed below the affected
There has been an increased interest in the use of grinding zone which are not visible [2].
advanced ceramic materials in the recent past due to their Low thermal coefficient of expansion, low density and
unique physical and mechanical properties. The advan- relatively high thermal conductivity are the special features
tages of ceramics over other materials include high hard- of silicon carbide ceramics. In view of these properties,
ness and strength at elevated temperatures, chemical SiC is expected to be used increasingly for heat resistant
stability, attractive high temperature wear resistance and parts [3]. Grinding is often the method of choice for mach-
low density [1]. Structural ceramics such as silicon ining ceramics in large-scale production and automation.
nitride and silicon carbide are now being increasingly Despite various research efforts in ceramic grinding over
used in bearings, valves, rotors and other applications the past two decades, much needs to be established to stan-
where a close dimensional tolerance is required. dardize models for process optimization for improving
Ceramic materials possess very low fracture tough- product quality and increasing productivity to reduce
ness compared with metals and alloys. This means that machining cost. The effective use of ceramics in industrial
they are very sensitive to the forces introduced due to applications demands the machining of ceramics with
machining. The major form of damage usually occurs as good surface finish and low surface damage. Therefore, in
surface and subsurface damage. The first type of damage the present work, an attempt has been made to examine
is due to radial cracks formed on the ground surface the effect of various process parameters and wheel proper-
ties on these responses during grinding of silicon carbide
with diamond abrasives. A genetic algorithm (GA) based
∗
Corresponding author. Tel.: +91-11-659-1443; fax: +91-11-868- optimization procedure has been developed to optimize
2037. grinding conditions, viz. depth of cut, work speed, grit size
E-mail address: [email protected] (P. Venkateswara Rao). and density, using a multi-objective function model.

0890-6955/$ - see front matter  2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0890-6955(03)00165-2
1328 A. Venu Gopal, P. Venkateswara Rao / International Journal of Machine Tools & Manufacture 43 (2003) 1327–1336

Nomenclature
C constant in mathematical model
C1 constant in surface roughness model
C2 constant in percentage area of surface damage model
d depth of cut, µm
f table feed rate, m/min
M grit size (mesh)
R grain density
Y machining response
%D percentage area of surface damage
Ra surface roughness measured, µm
d.o.f. degrees of freedom
MRR material removal rate, mm3/mm width/min
a depth of cut exponent in mathematical model
a1 depth of cut exponent in surface roughness model
a2 depth of cut exponent in percentage area of surface damage model
b feed rate exponent in mathematical model
b1 feed rate exponent in surface roughness model
b2 feed rate exponent in percentage area of surface damage model
g grit size exponent in mathematical model
g1 grit size exponent in surface roughness model
g2 grit size exponent in percentage area of surface damage model
h grain density exponent in mathematical model
h1 grain density exponent in surface roughness model
h2 grain density exponent in percentage area of surface damage model

2. Literature review men were investigated experimentally while grinding hot

pressed silicon nitride with resinoid bonded diamond
A major impediment to engineering applications of grinding wheels at different work speeds [5]. All finished
ceramics is their hardness and brittleness, which often surfaces showed surface damage over the range of wheel
render them difficult and costly to machine. Abrasive grit sizes employed. No loss of strength was reported for
machining of ceramics by means of grinding with dia- grinding in the longitudinal direction as the grit depth
mond wheels is the primary process used in achieving of cut increases. For transverse grinding, there was a
the desired tolerances and surface integrity. Grinding of decrease in strength as the grit depth of cut increases
silicon carbide is difficult because of its low fracture beyond a critical value of grit size. Modeling and optim-
toughness, making it very susceptible to cracking. The ization of the process could have further strengthened
grinding process is mostly conducted under moderate this work. Allor et al. [6] have investigated the effects
conditions requiring extensive machining. Efficient of diamond mesh size and concentration in the grinding
grinding of high performance ceramics requires wheel, wheel speed and the type of grinding on strength,
judicious selection of operating parameters to maximize surface roughness, power consumption, vertical force
removal rate while controlling surface integrity [4]. and wheel wear. The results showed that the strength of
Lowering grinding costs by using faster removal rates is the ceramic was affected by the diamond mesh size when
constrained mainly by surface damage to the ceramic creep-feed grinding was used, whereas diamond concen-
workpiece because of the median/lateral cracks that ema- tration and speed of the wheel were found to be signifi-
nate during grinding. So, it would be essential to under- cant factors when pendulum-feed grinding was perfor-
stand the mechanism of material removal and also to med.
evaluate the significance of the process parameters on Unfortunately, the ground ceramic components are
the required responses. most likely to contain a deformed layer,
Grinding is a very complicated process which surface/subsurface micro-cracks, phase transformation,
involves many parameters such as wheel speed, depth residual stresses and other types of damage. The per-
of cut, work speed, grit size, density, etc. The relation- formance and reliability of ceramic components are
ships of grit depth of cut and grind direction with influenced strongly by the damages introduced during
strength and surface characteristics of the ground speci- grinding. Malkin and Hwang [7] have studied and ana-
 A. Venu Gopal, P. Venkateswara Rao / International Journal of Machine Tools & Manufacture 43 (2003) 1327–1336 1329

lyzed the mechanism of material removal in ceramic roughness prediction model for turning mild steel using
grinding with the help of indentation fracture mechanics response surface methodology to give the factor effects
approach and the machining approach. It was shown of the individual process parameters. They have also
with the first approach that median/radial cracks are usu- attempted to optimize the surface roughness as an objec-
ally associated with strength degradation, and lateral tive function using GA.
cracks with material removal. Xu et al. [8] have demon- Most of the research has been conducted with silicon
strated that the mechanism of material removal and the nitride as workpiece material to investigate material
effect of machining induced damage on strength of removal rate, effect of residual stresses on fracture
advanced ceramics can be controlled by approximately strength and surface integrity in grinding. This paper
tailoring the microstructure. This not only promotes easy focuses on the effect of grit size, grit density, depth of
and well controlled material removal by grain dislodge- cut and work feed on surface finish and damage pro-
ment during machining, but also suppresses the forma- duced during grinding of silicon carbide, as this is one
tion of strength degrading cracks. of the materials being used increasingly in structural
Detailed knowledge on the effect of the grinding pro- applications. The results of the experimental work are
cess on surface integrity gives the opportunity for a bet- used to find the significance of the grinding variables
ter exploitation of ceramic materials by improved pro- on the machining responses by developing mathematical
cess conditions. Pfeiffer and Hollstein [9] have used the models. The mathematical models thus developed are
X-ray diffraction technique for determining the damage further utilized to find the optimum grinding parameters
induced in ground silicon nitride and alumina and ther- using genetic algorithms employing a multi-objective
eby established correlations between micro-plastic defor- function model.
mation and amount of damage. These investigations
show that in the case of lapped and ground alumina and
of ground silicon nitride, bending strength is dominated 3. Methodology
by machining induced damage. In the case of lapped sili-
con nitride, the effect of damage is compensated by To obtain applicable and practical predictive quanti-
machining induced compressive residual stresses. Dani- tative relationships, it is necessary to model the grinding
els [10] has investigated the influence of surface grinding responses and the process variables. These models would
parameters such as diamond abrasive type, wheel speed be of great use during optimization of the ceramic grind-
and down feed on the rupture strength of silicon carbide. ing process. The experimental results were used to model
It was found that more severe grinding conditions, such the various responses using the multiple regression
as higher normal forces and power consumption, did not method by using a non-linear fit among the response and
significantly reduce the mean rupture strength of the the corresponding significant parameters. The signifi-
material. The most encouraging aspect inferred from cance of the parameters on the responses could be found
these results was that grinding conditions could be using ANOVA. In this work, a commercially available
changed in order to optimize the process without sig- mathematical software package MATLAB was used for
nificant structural damage to the work material. the computation of regression constants and parameters.
Models contribute significantly to the comprehension The mathematical modeling of surface roughness and
of the process itself, and form the basis for the simul- surface damage in the grinding of ceramic material
ation of the grinding processes. They thus create a pre- involved lots of other factors, such as wheel speed, type
condition for increased efficiency while ensuring a high of abrasive grit, grit density, etc. However, to facilitate
product quality at the same time. Tonshoff et al. [11] experimental data collection, only four dominant factors
have described the state of the art in the modeling and were considered in the planning of the experimentation.
simulation of grinding processes, comparing the differ- A high precision hydraulic surface grinding machine was
ent approaches to modeling. Furthermore, the benefits as used to grind silicon carbide work material with diamond
well as the limits of model application and simulation grinding wheels at a speed of 36 m/s. The silicon carbide
were discussed in their paper. Liao and Chen [12] have material has been made from 3 µm particle sized powder
used back-propagation neural networks for modeling and at a sintering temperature of 2100 °C, which gives a
optimizing the creep-feed grinding of alumina with dia- density of 3.17 g/cm3 after processing. The factors con-
mond wheels. The authors have shown that it is possible sidered were depth of cut, feed rate, grit size, and grit
to obtain a global optimal solution by the proper use of density. The experiments were planned using a complete
the Boltzman factor. Jain and Jain [13] have also used 34 factorial design. Based on this, a total of 81 experi-
a similar approach for optimum selection of machining ments, each having a combination of different levels of
conditions in abrasive flow machining using neural net- factors as shown in Table 1, were carried out.
works. The results have been validated by comparing the The responses measured were tangential grinding
optimized machining conditions obtained using genetic force, surface roughness and percentage area of surface
algorithms. Suresh et al. [14] have developed a surface damaged. The grinding force was measured with an
1330 A. Venu Gopal, P. Venkateswara Rao / International Journal of Machine Tools & Manufacture 43 (2003) 1327–1336

Table 1 dom search. The mechanics of GA is simple, involving

Values of test variables copying of binary strings. Simplicity of operation and
computational efficiency are the two main attractions of
Variables Values of different levels
the genetic algorithmic approach. The computations are
Designation Description Low Medium High carried out in three stages to get a result in one gener-
(⫺) (0) (+) ation or iteration. The three stages are reproduction,
crossover and mutation.
d Depth of cut (µm) 5 15 30 In order to use GA to solve any problem, the variable
f Feed rate (m/min) 5 10 15 is typically encoded into a string (binary coding) or
M Grit size (mesh) 120 240 500 chromosome structure which represents a possible sol-
R Grit density 50 75 100
ution to the given problem. GA begin with a population
of strings (individuals) created at random. The fitness of
AMTI make MC36 strain gauge type dynamometer. The each individual string is evaluated with respect to the
surface roughness was measured with Talysurf-6 at 0.8 given objective function. Then this initial population is
mm cut-off value. The amount of damage was measured operated on by three main operators—reproduction,
with the optical microscope, a Leica TCS SP2 system, crossover, and mutation—to create hopefully a better
at 100 magnification. population. Highly fit individuals or solutions are given
opportunities to reproduce by exchanging pieces of their
3.1. Mathematical formulation genetic information, in the crossover procedure, with
other highly fit individuals. This produces new “off-
Multiple regression analysis is practical, economical spring” solutions, which share some characteristics taken
and relatively easy for use and it was widely used for from both the parents. Mutation is often applied after
modeling and analyzing experimental results [15]. The crossover by altering some genes (i.e. bits) in the off-
mathematical models for the ceramic grinding with the spring. The offspring can either replace the whole popu-
parameters under consideration are represented by: lation (generational approach) or replace less fit individ-
uals (steady-state approach). This new population is
Y ⫽ f(d,f,M,R) (1) further evaluated and tested for some termination cri-
where Y is the grinding response, f is the response func- teria. The reproduction-crossover-mutation-evaluation
tion, and d, f, M, R are grinding variables. Expressed in cycle is repeated until the termination criteria are met.
non-linear form, Eq. (1) becomes
Y ⫽ Cdaf bMgRh (2) 4. Results and discussion
In this work, the following mathematical models were
formulated: The variation of tangential grinding force and surface
roughness with respect to the process variables selected
Surface roughness model: Ra ⫽ C1da1f b1Mg1Rh1 (3) in the present study is shown graphically in Figs. 1 and
Surface damage model: %D ⫽ C2da2f b2Mg2Rh2 (4) 2. Some of the typical observations of the percentage
area of surface damage are given in Fig. 3. It is observed
To facilitate the determination of constants and para- from the results shown in Figs. 1 and 2 that the tangen-
meters, these mathematical models were linearized by tial grinding force and surface roughness decrease with
performing a logarithm transformation. The above func- a decrease in depth of cut, feed rate and grit density and
tion can be represented in linear mathematical form as an increase in the grit size. The results comply with the
follows: trends available in the literature [1,2,4–6].
lnY ⫽ lnC ⫹ alnf ⫹ blnd ⫹ glnM ⫹ hlnR (5)
4.1. Roughness model
The constants and parameters C, a, b, g and h can
then be solved by using multiple regression analysis with The dimensional accuracy and surface finish of any
the help of experimental results. manufacturing process have become critical because of
increased quality demands. There are various factors that
3.2. Optimization using genetic algorithm govern surface finish in grinding and hence, the develop-
ment of analytical or empirical models for reliable pre-
Genetic algorithms (GA) form a class of adaptive heu- diction of machining performance becomes a key issue.
ristics based on principles derived from the dynamics of ANOVA has been performed to find the effect of fac-
natural population genetics [16]. The searching process tors and their interactions on the responses [17]. Though
simulates the natural evaluation of biological creatures the experiments were conducted using full factorial
and turns out to be an intelligent exploitation of a ran- design, replication of experiments with each combi-
A. Venu Gopal, P. Venkateswara Rao / International Journal of Machine Tools & Manufacture 43 (2003) 1327–1336 1331

Fig. 1. Variation of tangential grinding force with grinding variables.

Fig. 2. Variation of surface roughness with grinding variables.

1332 A. Venu Gopal, P. Venkateswara Rao / International Journal of Machine Tools & Manufacture 43 (2003) 1327–1336

Fig.(a) d:5µm; f:15m/min; M:500; R:100 and %D:1.38, (b) d:5µm; f:5m/min; M:120; R:50 and %D:3.07, (c) d:15µm; f:15m/min; M:120; R:75 and %D:3.21,
(d) d:5µm; f:15m/min; M:240; R:50 and %D:1.61, (e) d:30µm; f:15m/min; M:500; R:100 and %D:2.34, (f) d:30µm; f:15m/min; M:120; R:75 and %D:3.92
(magnification is same for all the figures shown in Fig.(a))

Fig. 3. Surface damage observed under optical microscope.

nation could not be carried out due to the limitation of Due to limitation of resources, testing could be carried
experimental resources. Accordingly, it has been out on only 27 specimens. Since all the factors had three
assumed that the four factor interaction is not present levels, an L27 orthogonal array (OA) has been used to
and the corresponding sum of squares and degrees of select the specimens for testing. The results of percent-
freedom have been taken as residual/error to conduct age area of surface damage are shown in Table 3. How-
the ANOVA. ever, since the experiments involved four factors (d, f,
The results of ANOVA for surface roughness, shown M and R) only, these factors are assigned to columns 1,
in Table 2, indicate that the variables d, f, M and R are 5, 8 and 11, respectively, of L27 OA. This assignment
significant at the 99% confidence level and their interac- makes it possible to test the significance of not only the
tions are not significant even at the 95% confidence factors but also the interactions between “d and f”, “d
level. Thus, the parameters d, f, M and R are considered and M” and “d and R” for the levels of factors con-
while developing the model for roughness, and it is sidered.
given by: The results of ANOVA for the percentage area of sur-
face damage, shown in Table 4, indicate that the vari-
Ra ⫽ 0.86d0.1843f 0.5253M⫺0.2866R⫺0.2444 (6)
ables d, f and M are only significant at the 99% confi-
dence level, whereas the parameter R and the interactions
4.2. Surface damage model
are not significant even at the 95% confidence level.
Thus, only the parameters d, f and M have been con-
The measurement of machining induced damage in
sidered in developing the percentage area of surface
ceramics is of interest for two reasons. First, the damage
damage model, and it is given by:
induced will influence the service behavior of the mech-
anical parts. The effect of damage on strength is parti- %D ⫽ 24.44d0.2857f⫺0.3M⫺0.4140 (7)
cularly important in brittle materials, like ceramics.
Second, the damage induced is a consequence of forces It can be observed from the above models that the
occurring during machining. In order to understand the surface roughness decreases with a decrease in depth of
influence of various factors on the induced grinding cut and feed and with an increase in grit number as well
damage, the development of analytical or empirical as grit density. It is also observed that decrease in depth
models for reliable prediction of machining performance of cut or an increase in feed or grit number would
becomes another key issue. decrease the percentage area of surface damage. The
 A. Venu Gopal, P. Venkateswara Rao / International Journal of Machine Tools & Manufacture 43 (2003) 1327–1336 1333

Table 2
ANOVA for surface roughness

Source d.o.f. Sum of squares Mean squares Fcal F0.05 F0.01

d 2 0.1585 0.07925 136.34 3.63 7.54 Significant at

99%
f 2 0.4733 0.23665 407.14
M 2 0.2339 0.11695 201.20
R 2 0.0433 0.010825 37.25

df 4 0.0009 0.000225 0.387 3.01 Not significant

even at 95%
confidence level
dM 4 0.0019 0.000475 0.817
dR 4 0.0012 0.0003 0.516
fM 4 0.0057 0.001425 2.45
fR 4 0.0009 0.000225 0.387
MR 4 0.0022 0.00055 0.946

dfM 8 0.0036 0.00045 0.774 2.59 Not significant

even at 95%
confidence level
dfR 8 0.0023 0.0002875 0.495
dMR 8 0.0020 0.00025 0.430
fMR 8 0.0079 0.0009875 1.700

Residual/error 16 0.0093 0.00058125 –

Total 80 0.9469

major finding of this study is that the percentage area of

Table 3 surface damage decreases with an increase in feed within
Experimental results for percentage area of surface damage the range considered. This could be due to the less time
available for the crack to propagate at higher feed rates.
L-27 d f M R %D This provides an opportunity to maximize the material
removal rate by selecting the proper process parameters.
1 5 5 120 50 3.07
2 5 10 240 75 2.11
Hence, in order to achieve better surface finish and mini-
3 5 15 500 100 1.38 mum damage with a better material removal rate, a pro-
4 5 5 240 100 2.53 per combination of depth of cut, feed, grit size, and grit
5 5 10 500 50 1.52 density must be selected within their operating range.
6 5 15 120 75 2.41 So, it was decided to optimize the grinding process for
7 5 5 500 75 1.83
8 5 10 120 100 2.84
maximization of material removal rate with surface fin-
9 5 15 240 50 1.61 ish and damage as constraints, thus forming a multi-
10 15 5 120 50 4.12 objective function.
11 15 10 240 75 2.71
12 15 15 500 100 1.79 4.3. Optimization of ceramic grinding
13 15 5 240 100 3.32
14 15 10 500 50 1.94
15 15 15 120 75 3.21
Optimization of machining parameters increases not
16 15 5 500 75 2.56 only the utility for machining economics, but also the
17 15 10 120 100 3.72 product quality to a great extent. In this context, an effort
18 15 15 240 50 2.32 has been made to estimate the surface roughness and
19 30 5 120 50 6.19 percentage area of surface damage using experimental
20 30 10 240 75 3.33
21 30 15 500 100 2.34
data. It has also been attempted to optimize the grinding
22 30 5 240 100 4.15 process using GA in order to achieve good surface finish
23 30 10 500 50 2.42 and minimum damage.
24 30 15 120 75 3.92 A simple GA code was used in the present study. The
25 30 5 500 75 2.91 procedure involved in the optimization, using GA, of the
26 30 10 120 100 4.59
27 30 15 240 50 2.85
silicon carbide ceramic grinding process is shown in Fig.
4. The problem of optimization of the ceramic grinding
1334 A. Venu Gopal, P. Venkateswara Rao / International Journal of Machine Tools & Manufacture 43 (2003) 1327–1336

Table 4
ANOVA for percentage area of surface damage

Source d.o.f. Sum of squares Mean squares Fcal F0.05 F0.01

d 2 09.98 4.990 62.38 5.14 10.92 Significant at 99%

confidence level
f 2 04.43 2.215 27.69
M 2 13.29 6.645 83.06

R 2 00.15 0.075 00.94 5.14 Not significant even at

95% confidence level
df 4 00.49 0.1225 01.53 4.53
dM 4 01.07 0.2675 03.34
dR 4 00.24 0.06 00.75

Residual/error 6 00.48 0.08 –

Raⱕ(Ra)max,
%Dⱕ(%D)max and
xliⱕxiⱕxui
MRR ⫽ fd
Ra ⫽ 0.86(d)0.1843(f)0.5253(M)⫺0.2866(R)⫺0.2444
%D ⫽ 24.44(d)0.2857(f)⫺0.3(M)⫺0.4140
5 µmⱕdⱕ30 µm
5 m / minⱕfⱕ15 m / min
120ⱕDⱕ500
50ⱕRⱕ100
where xli and xui are the lower and upper bounds on pro-
cess variables xi, and (Ra)max and (%D)max are the
maximum allowable values of surface roughness and
number of flaws.
Fig. 4. General procedure of GA. The GA program had been written in the MATLAB
environment. The maximizing function was written faci-
litating the user to set the constraints. The optimization
process can be described as maximizing the MRR sub- was carried out for different values of constraints on sur-
ject to a criteria set of constraints on surface roughness, face roughness (0.16–0.4 µm) and percentage area of
percentage area of surface damage and input variables. surface damage (1.6–4%). Some typical results, obtained
In order to optimize the present problem using GA: by GA for different surface roughness values imposing
a constraint on percentage area of surface damage as 2,
(a) the following parameters were specified by practice, are presented in Table 5. In a similar way, another set
to get optimal solutions with low computational of typical results, obtained by GA for different percent-
effort: age area of surface damage values placing a constraint
on surface roughness as 0.25 µm, are presented in
앫 maximum number of generations = 300, Table 6.
앫 total string length = 30, It can be observed from the results of optimization
앫 mutation probability = 0.1 and that the MRR increases by almost 10 times by increasing
앫 crossover probability = 0.03; the roughness constraint from 0.16 to 0.4 µm at 2% area
(b) the constrained optimization problem is stated as fol- of surface damage, whereas it increases only 2.5 times
lows: by increasing the damage constraint from 1.6% to 3.0%
at a roughness value of 0.25 µm and remains constant
앫 maximize MRR subject to thereafter. This shows that no improvement in MRR can
 A. Venu Gopal, P. Venkateswara Rao / International Journal of Machine Tools & Manufacture 43 (2003) 1327–1336 1335

Table 5
Optimization results of GA at various values of Ra and at 2% area of surface damage

Constraints Output of GA Values of constraints using output of GA

(Ra)max (µm) (%D)max d (µm) f (m/min) M R Ra (µm) (%D) MRR (mm3/mm

width/min)

0.155 2 6.4784 5.0014 497.8295 99.6999 0.1548 1.9629 32.4007

0.175 2 7.9673 5.8201 499.9083 99.3983 0.1741 1.9864 46.3708
0.200 2 9.0321 7.2004 498.8107 99.4481 0.1993 1.9864 65.0340
0.225 2 10.7470 8.4640 492.7791 99.4934 0.2248 1.9989 90.9631
0.250 2 12.6997 9.7158 498.6713 97.7521 0.2495 1.9989 123.388
0.275 2 14.4083 11.2128 494.6621 99.5517 0.2747 1.9946 161.5570
0.300 2 16.5146 12.5864 499.6223 99.4741 0.2986 1.9948 207.8587
0.325 2 17.4539 14.4231 494.9530 99.6562 0.3247 1.9533 251.7402
0.350 2 20.2663 14.6675 497.2222 85.2220 0.3495 1.9994 297.2558
0.375 2 20.9556 14.9883 494.2799 72.1604 0.3710 1.9841 314.0886
0.400 2 22.8495 14.8057 496.9494 55.0278 0.3996 1.9933 338.3029

Table 6
Optimization results of GA at various percentage area of surface damages and at Ra = 0.25 µm

Constraints Output of GA Values of constraints using output of GA

(Ra)max (µm) (%D)max d (µm) f (m/min) M R Ra (µm) (%D) MRR (mm3/mm

width/min)

0.25 1.50 6.4543 12.3946 495.0068 99.2775 0.2499 1.4970 79.9979

0.25 1.75 9.4716 10.6927 498.2692 96.8739 0.2492 1.7413 101.2776
0.25 2.00 12.6997 9.7158 498.6713 97.7521 0.2495 1.9989 123.3880
0.25 2.25 16.7058 8.8068 491.2927 99.7087 0.2491 2.2437 147.1243
0.25 2.50 22.4213 8.0192 499.7077 99.6853 0.2491 2.4924 179.8007
0.25 2.75 28.3410 7.3815 498.4317 98.8008 0.2498 2.7330 209.1995
0.25 3.00 29.9254 7.2185 496.6276 98.3001 0.2499 2.7975 216.0157
0.25 3.25 29.9254 7.2185 496.6276 98.3001 0.2499 2.7975 216.0157
0.25 3.50 29.9254 7.2185 496.6276 98.3001 0.2499 2.7975 216.0157
0.25 3.75 29.9254 7.2185 496.6276 98.3001 0.2499 2.7975 216.0157
0.25 4.00 29.9254 7.2185 496.6276 98.3001 0.2499 2.7975 216.0157

be seen even by relaxing the constraint on surface dam- 5. Conclusions

age beyond 3%. It can also be observed that higher grit
densities were required for all the values of surface dam- The investigations of this study indicate that the para-
age constraints for a good surface finish, leading to an meters feed rate, depth of cut and grit size are the pri-
increase in the cost of the grinding process. On the other mary influencing factors which affect the surface integ-
hand, the requirement on grit density decreases as the rity of silicon carbide during grinding. They also indicate
constraint on surface roughness increases even at low that the percentage area of surface damage decreases
damage. Hence, it can be concluded from the above opti- with an increase in feed rate and is little affected by the
mization results that the material removal rate and the variation in grit density, within the range considered.
cost of grinding are influenced more by the surface The approach presented in this paper provides an
roughness constraint than by that of surface damage. So, impetus to develop analytical models, based on the
the cost of grinding ceramics can be brought down con- experimental results, to predict the general trends of
siderably by proper selection of grinding parameters ground workpiece surface roughness and percentage area
with the help of the optimization procedure developed of surface damage in terms of the significant parameters
in this work. under consideration. Optimal grinding conditions are
1336 A. Venu Gopal, P. Venkateswara Rao / International Journal of Machine Tools & Manufacture 43 (2003) 1327–1336

also obtained for maximization of material removal formation mechanisms in grinding silicon nitride, Journal of
using surface roughness and percentage damage, com- Materials Research 11 (7) (1996) 1717–1724.
[5] J.E. Mayer Jr., G.P. Fang, Effect of grinding parameters on sur-
puted by the models developed, as constraints. The face finish of ground ceramics, Annals of the CIRP 44 (1) (1995)
results of optimization conclude that the material 279–282.
removal and the cost of grinding are influenced more [6] R.L. Allor, T.J. Whalen, J.R. Baer, K.V. Kumar, Machining of
by the constraint on surface roughness than by surface silicon nitride: experimental determination of process/property
damage. The optimized model developed offers a sol- relationships, in: Proceedings of International Conference on
Machining Advanced Materials, NIST SP847, 1993, pp. 223–234.
ution to reduce the cost of machining, thereby making [7] S. Malkin, T.W. Hwang, Grinding mechanisms for ceramics,
silicon carbide a more commercially viable material for Annals of the CIRP 45 (2) (1996) 569–580.
industrial applications. [8] H.H.K. Xu, N.P. Padutre, S. Jahanmir, Effect of microstructure
on material removal mechanisms and damage tolerance in abras-
ive machining of silicon carbide, Journal of the American Cer-
amic Society 78 (9) (1995) 2443–2448.
Acknowledgements [9] W. Pfeiffer, T. Hollstein, Damage determination and strength pre-
diction of machined ceramics by X-ray diffraction techniques, in:
Proceedings of Conference on Experimental Mechanics, Copen-
The authors would like to express their sincere thanks hagen, Denmark, 1990, pp. 235–245.
to the Department of Science and Technology, New [10] W.H. Daniels, Superabrasives for ceramic grinding and finishing,
Delhi, India, for financially supporting this research pro- SME Technical Paper EM89-125, 1989.
ject and also to the International Advanced Research [11] H.K. Tonshoff, J. Peters, I. Inasaki, T. Paul, Modelling and simul-
Center for Powder Metallurgy and New Materials, Hyd- ation of grinding processes, Annals of the CIRP 41 (2) (1992)
677–688.
erabad, India, for supplying the work material. [12] T.W. Liao, L.J. Chen, A neural network approach for grinding
processes: modelling and optimization, International Journal of
Machine Tools and Manufacture 34 (7) (1994) 919–937.
[13] R.K. Jain, V.K. Jain, Optimum selection of machining conditions
References in abrasive flow using neural networks, Journal of Materials Pro-
cessing Technology 108 (2000) 62–67.
[1] B.P. Bandyopadhyay, The effects of grinding parameters on the [14] P.V.S. Suresh, P. Venkateswara Rao, S.G. Deshmukh, A genetic
strength and surface finish of two silicon nitride ceramics, Journal algorithmic approach for optimization of surface roughness pre-
of Materials Processing Technology 53 (1995) 533–543. diction model, International Journal of Machine Tools and Manu-
[2] T.M.A. Maksoud, A.A. Maksoud, A.A. Mokbel, J.E. Morgan, facture 42 (2002) 675–680.
Evaluation of surface and sub-surface cracks of ground ceramic, [15] E.J.A. Armarego, R.H. Brown, in: The Machining of Metals,
Journal of Materials Processing and Technology 88 (1999) Prentice Hall, New Jersey, 1984, pp. 254–291.
222–243. [16] D.E. Goldberg, Genetic Algorithms, Addison Wesley Longman
[3] I. Inasaki, Grinding of hard and brittle materials, Annals of the Inc, India, 1999.
CIRP 36 (2) (1987) 463–471. [17] D.C. Montgomery, Design and Analysis of Experiments, John
[4] H.H.K. Xu, S. Jahanmir, L.K. Ives, Material removal and damage Wiley and Sons (Asia) Pvt. Ltd, Singapore, 2001.